codex-history
Pass
Audited by Gen Agent Trust Hub on May 17, 2026
Risk Level: SAFE
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to locate and read local files including
~/.codex/history.jsonl,~/.codex/state_5.sqlite, and session-specific JSONL logs to reconstruct past workflows. - [DATA_EXPOSURE]: Accesses local application data consisting of user conversation history and metadata (like
cwd). This data access is the primary intended function of the skill and is required to fulfill the user's request to resume past Codex sessions. - [PROMPT_INJECTION]: The skill presents an indirect prompt injection surface by ingesting historical logs which may contain untrusted content from previous LLM outputs or user inputs.
- Ingestion points:
history.jsonl, session JSONL logs, andstate_5.sqlitedatabase file. - Boundary markers: None specified in the instructions.
- Capability inventory: File reading, data extraction, and task summarization/resumption.
- Sanitization: No explicit sanitization or validation of historical content is mentioned.
Audit Metadata